Title: Signaling in Bayesian Games
Speaker: | Chaitanya Swamy |
Affiliation: | University of Waterloo |
Room: | MC 5501 |
Abstract:
We
study
the
optimization
problem
faced
by
an
informed
principal
in
a
Bayesian
game,
who
can
reveal
some
information
about
the
underlying
random
state
of
nature
to
the
players
(thereby
influencing
their
payoffs)
so
as
to
obtain
a
desirable
equilibrium.
This
yields
the
following
signaling
problem:
what
information
should
the
principal
reveal
to
achieve
his
goal?
This
is
a
natural
design
question
motivated
by
uncertainty
in
games
and
has
attracted
much
recent
attention.
I
will
highlight
some
recent
almost-optimal
hardness
results
and
some
approximation
algorithms
for
Bayesian
two-player
zero-sum
games
and
Bayesian
network
routing
games.
Both
these
classes
admit
a
canonical,
tractable
choice
of
equilibrium,
which
also
decouples
the
concerns
of
optimal-signaling
computation
and
equilibrium
computation.
For
Bayesian
zero-sum
games,
wherein
the
principal
seeks
to
maximize
the
equilibrium
utility
of
a
player,
we
exploit
duality
and
the
equivalence
of
optimization
and
separation
to
obtain
hardness
results
and
algorithms.
No background in game theory will be assumed.